By Steve Merrill, Founder of WRKNG Digital | July 29, 2026
The five feed fields that decide whether AI recommends your products are the title, attributes, GTIN, availability and price, and the structured description. Get these right and AI can read your catalog. Get them wrong and you're invisible. Here's each one.
People think AI visibility is some deep technical mystery. It's mostly five fields. We audited 2,400 products last year and only 11% had these done well. Fix them and you're ahead of almost everyone.
1. Product Title
The title is the first thing AI reads. Vague titles like "Classic Tee" tell a machine nothing. Specific titles like "Men's Organic Cotton Crew T-Shirt, Navy" match real shopper questions. Put the key attributes right in the title.
2. Product Attributes
Attributes are how AI matches a product to intent. Color, size, material, use case, and more. Google's product attributes documentation lists the fields that matter. Blank attributes are the most common reason a product never gets recommended.
3. GTIN and Unique Identifiers
GTINs and other unique IDs tell AI and shopping systems exactly which product you're selling. They tie your listing to a known item. Google's product data spec treats these as core fields. Missing IDs make your catalog harder to trust.
4. Availability and Price
AI won't confidently recommend a product it thinks is out of stock or wrongly priced. Keep availability and price accurate and synced. Schema.org's Offer type carries this data in a format machines read directly. Stale values quietly kill recommendations.
5. Structured Description
A description written for AI reads like clear answers, not marketing fluff. Name the specs. State the use cases. Keep it specific. This is the text AI extracts and quotes, so vague copy leaves you with nothing to be cited for.
How We Chose These Fields
We ranked the five fields that most often decide whether a product gets recommended, based on what we see fixing real Shopify catalogs. Each one is a direct, measurable lever.
FAQ
Which product feed field matters most for AI?
The title, closely followed by attributes. AI reads the title first to understand what you sell, then uses attributes to match it to a shopper's question.
Do I really need GTINs for AI recommendations?
They help a lot. Unique identifiers tie your listing to a known product, which makes AI and shopping systems trust and place it more confidently.
How should I write a product description for AI?
Write clear answers, not fluff. Name the specs and use cases in specific language. AI extracts and quotes real detail, not vague marketing copy.
How do I check my feed fields?
Review your feed for blank attributes, vague titles, missing GTINs, and stale price or stock. Then test an AI model to see if products in that category surface your store.
Want to see how your feed fields score for AI readability? Get your product feed AI audit here.
